|
#include "mpi.h"
#include <stdio.h>
#include <math.h>
double f(double);
double f(double x)
/* 定义函数f(x) */
{
return (4.0 / (1.0 + x*x));
}
int main(int argc,char *argv[])
{
int done = 0, n, myid, numprocs, i;
double PI25DT = 3.141592653589793238462643;
/* 先给出已知的较为准确的p值*/
double mypi, pi, h, sum, x;
double startwtime = 0.0, endwtime;
int namelen;
char processor_name[MPI_MAX_PROCESSOR_NAME];
MPI_Init(&argc,&argv);
MPI_Comm_size(MPI_COMM_WORLD,&numprocs);
MPI_Comm_rank(MPI_COMM_WORLD,&myid);
MPI_Get_processor_name(processor_name,&namelen);
fprintf(stdout,"Process %d of %d on %s\n",
myid, numprocs, processor_name);
n = 0；
if (myid == 0)
{
printf("Please give N=");
scanf(&n);
startwtime = MPI_Wtime();
}
MPI_Bcast(&n, 1, MPI_INT, 0, MPI_COMM_WORLD);/*将n值广播出去*/
h = 1.0 / (double) n;/*得到矩形的宽度，所有矩形的宽度都相同*/
sum = 0.0; /*给矩形面积赋初值*/
for (i = myid + 1; i <= n; i += numprocs)
/*每一个进程计算一部分矩形的面积，若进程总数numprocs为4，将0-1区间划分为100个矩形，则各个进程分别计算矩形块
0进程 1，5，9， 13，...，97
1进程 2，6，10，14，...，98
2进程 3，7，11，15，...，99
3进程 4，8，12，16，...，100
*/
{
x = h * ((double)i - 0.5);
sum += f(x);
}
mypi = h * sum;/*各个进程并行计算得到的部分和*/
MPI_Reduce(&mypi, &pi, 1, MPI_DOUBLE,
MPI_SUM, 0,
MPI_COMM_WORLD);
/*将部分和累加得到所有矩形的面积，该面积和即为近似p值*/
if (myid == 0)
/*执行累加的0号进程将近似值打印出来*/
{
printf("pi is approximately %.16f, Error is %.16f\n",
pi, fabs(pi - PI25DT));
endwtime = MPI_Wtime();
printf("wall clock time = %f\n", endwtime-startwtime);
fflush( stdout );
}
MPI_Finalize();
}
|
例子